LINGUIST List 2.161

Wednesday, 24 Apr 1991

Disc: Morphophonology (Part 2)

Editor for this issue: <>

Directory

1. , Phonology (Re: Hammond on English Stress)
2. Harry Bochner, Morphophonology
3. John Coleman, RE: Morphophonology
4. , Phonology
5. "ELISE EMERSON MORSE-GAGNE", orthography and phonological intuitions

I may not have made one thing clear, since the argumentation gets
involved. Here goes again:

 If, as Mike says, the stress pattern of Orchestra is
 the predictable one, and that of canAsta is due to
 the same kind of exception marking that kentUcky has,
 then Industry goes with Orchestra, and there is then
 neither any need nor any DISTRIBUTIONAL analogy for
 attributing a final /y/ rather than /I/ or /i/ to
 Industry (as usual, I mark primary stress by capitalization).

I may be wrong, but at least I think this time I am clear.
--------------------------------------------------
Also, there may be another example like diabEtes, namely,
commIttee. Now, for me (like most American speakers?),
this rhymes exactly with kItty, so one could perhaps argue
that the final vowel is underlyingly (with the usual caveats)
/I/ rather than /i/. What is needed is a speaker who has a contrast
between final [i] and [I], who has [i] in committee (but [I] in words
like Kitty, and who also flaps his t's, in particular in committee.
I wonder if anyone reading this is, or knows, such a speaker?

Mike Hammond writes in Vol. 2, No. 156.
> I agree that there are a number of words like _Kentucky_ that would be
> treated as exceptions if, as I am arguing, English stress is
> rule-governed.
....
> Alexis likens English to Greek where stress is restricted to the last
> three syllables, but, he maintains, is lexical in that domain.
....
> The moral is that even though stresss is "lexical" in the
> three-syllable domain, it would be a mistake to say it's not
> rule-governed.

I'd like to summarize the point of disagreement as I understand it:
We agree that there are significant generalizations about English stress in
various morphological environments, and that are exceptional words,
particularly among underived words.

The disagreement comes, I think, from the fact that Mike (and SPE-style
phonology generally) seems to think that there is a simple dichotomy: either
a) stress is 'rule-governed' in the sense that every case can be ground out by
 the rule system (using exception features if necessary), or 
b) it is completely arbitrary lexically, not governed by any significant
 generalizations.

While I'm not sure what Alexis has in mind, I doubt that it's (b). In any
case, I claim that this dichotomy is a false one.

What we need is a theory that allows us to express generalizations about
listed facts. This is what word-based theories of Morphology are about:
when we say that 'sanity' has a lexical entry in such a theory, we are
_not_ denying the validity of the generalization that -ity often forms
nouns from adjectives. Similarly, saying that stress is lexical in English
does not deny that it is often predictable.

This interacts with the phonology that Alexis and Mike have been discussing
in the following way: if there were a well-motivated abstract analysis that
accounted for stress in all English words, using exception features only for
plausible 'phonetic idioms' and unassimilated loans, I for one would gladly
accept it. But I agree with Alexis in rejecting poorly motivated analyses
(like non-syllabic +y in 'industry') that do little more than increase
slightly the proportion of lexical items whose stress can be claimed to be
'rule-governed'. In a theory where saying "stress is unpredictable in this
word" does not deny the validity of the generalizations that hold in many
other cases, the logic of Mike's argument fails to go through. And I would
argue that such a theory is required for many reasons independent of these
immediate issues.

-- Harry Bochner
-- bochnerdas.harvard.edu

Mike Hammond writes in response to Alexis Manaster-Ramer's observation
that "you can do anything you want once you are allowed to take impossible
sound sequences and put them in your URs. If there are alternations, 
that is one thing. But in this case there are not." with two responses:
1. "there are bizillions of analyses out there ... where such 
representations are of course unpronounceable and contain impossible 
sequences of sounds." 2. "I don't understand why alternations are criterial".

But Alexis's objection is quite right, irrespective of how many
analyses fall foul of it. They could all be wrong; in fact I think
they mostly are precisely because of Alexis's logical objection.
(That doesn't mean they're all wholly wrong and couldn't be repaired,
though. They may only fall foul of "surface-trueness" for some quite
unnecessary reason.) As for 2, I didn't interpret Alexis to mean that
alternation is the ONLY criterion to consider in setting up abstract
phonological representations. But I think most people would agree that
support for an analysis from alternations is preferable to just
distributional support. Isn't it more-or-less the same as the
explanatory vs. descriptive distinction, respectively?

--- John Coleman

 I knew I wouldn't be able to stay out of the English stress
pattern debate for very long. It seems to me perfectly respect-
able linguists have argued that there are cases where the assign-
ment of stress in English is lexical but not exceptional. Lisa
Selkirk suggested (LI 11.3: 595), on the basis of words like
`catamaran', that stress *had* to be lexically marked, because
perfectly plausible */caTAmaran/ and */cataMAran/ had to be
prevented (the non-reduced vowel in the ultima rules out the cute
trick of an underlying final non-syllabic sonorant). In this
case *only* lexical marking of stress (or foot structure, which
is what Selkirk used for the same effect) will work because the
word is not usefully distinct from, say, DeCAmeron or AthaBAscan.
I would be really uneasy if we decided that `catamaran' and
`Athabascan' were `irregular' and `Decameron' was regular--there
is no historical or psycholinguistic justfication for such a
choice, and unlike *real* distribution facts (such as those deal-
ing with allophones, or even stress in languages like Polish,
which is much closer to being predictable) there are just too
many exceptions that don't *feel* foreign.
 Admittedly, Selkirk claims that this does not mean that
stress is unpredictable, only that `in certain cases options are
made available'. But this is (to me) like being a little bit
pregnant. If we have to learn how to stress this word, why
shouldn't we have to learn how to stress `cannister'?
 Secondly, isn't it much more likely that a child will find
the task of learning the stress for each word much simpler than
the task of reconstructing an underlying representation such as
/Ind5strj/, with a consonant sequence that grossly violates all
known consonant sequence constraints (note how much more bizarre
the form looks if we use the IPA rather than the American trans-
cription system--a fact that I suspect is not irrelevant). While
I agree that underlying representations *could* be unpronounce-
able (and in fact, Natural Phonologists argue that allophonic
rules have the *function* of rendering underlying forms pro-
nounceable), surely it is very unlikely that a learner will cling
to a pattern that already has lots of holes in it if that means
the same learner will be forced to reconstruct underlying forms
that doubly violate the sonority hierarchy.
 Lastly, psycholinguistic evidence (quite old now) on the Tip
of the Tongue phenomenon indicates that (at least for English
speakers) we know the stress pattern of a word even when we are
unable to retrieve it fully. If the notion of lexical vs. rule-
governed has any psychological reality at all, surely we could
not have access to stress patterns without being able to access
the segmental information, *unless* the stress pattern were
stored, not generated `on-the-fly'.

GEOFF NATHAN <GA3662SIUCVMB.BITNET>
SOUTHERN ILLINOIS UNIVERSITY AT CARBONDALE
CARBONDALE, IL, 62901, USA

In February a Penn graduate student named Katya Zubritskaya gave a talk
at the Penn Linguistics Colloquium on a study she had made of some Russian
language games. The study involved only children who were native speakers
of Russian; some were resident in the USSR, some in the States. Now, I
can't find my handout, so this should NOT be taken as exactly what Katya
said, but here goes. I think she found that, for one thing, the children
did not by any means agree with one another on how to syllabify the words
in question (I think syllabification was at issue), and for another thing,
that the ability to apply the rules of the game rapidly, speaking the
derived language with fluency, was linked to such cognitive abilities as
reading skills--whether the child read in English or Russian. More relevant
to the discssion on Russian orthography, however, was her finding that how
the "phonology" worked for the children appeared to depend on WHETHER THEY
READ RUSSIAN. The ones in this country have not learned much Russian
orthography, but according to Katya their speaking ability is perfectly
native. And their decisions about certain alterations in words which are
supposedly purely phonological in nature apparently differed from the
decisions made by the children who had learned to read and write Russian.
 Again, I am afraid I may not be reporting this very accurately,
but I found the talk interesting. The conference proceedings are published
every spring, probably about this time, and if Katya wrote up her paper for
the printed version it can be obtained from the Dept. of Linguistics,
Williams Hall, University of Pennsylvania, Philadelphia, PA 19104 for a
small sum like $5 or so, I think. Or anyone who is interested could just
write to Katya at that address. I don't think she has an e-mail address.
--Elise Morse-Gagne

[End Linguist List, Vol. 2, No. 161]